Description
This Rmd file investigates residence time of individual fish in
different reaches. This will help us determine in what time window
covariates are most influential to fish movement decisions.
Covariate inclusion in Siegel et al. 2021: “Environmentally
triggered shifts in steelhead migration behavior and consequences for
survival in the mid-Columbia River”

“For each population group, we fit a 3-component mixture model [32]
to log-transformed travel times using the package mixtools [33] to
assign a probability of migration behavior to each fish that was
detected at both dams (n = 32,740).”
“We considered the following environmental variables in models for
the probability of delay: river temperature (T), river flow (F), dam
spill volume (S), and the date of first detection at Bonneville Dam (D).
The environmental variables (T, F, and S), which were accessed from the
Columbia River Data Access in Real Time database [36], were averaged
over the median fast fish travel time period (8 days) starting on the
day of first detection at Bonneville Dam. These 8 days represent the
period of exposure during which a fish makes the “decision” to delay or
to continue upstream.”
Proposed covariate structure
Continuous
- Temperature
- In mainstem sites, use the temperature value at the median residence
time of “fast” fish - the fish that move straight through.
- Use temperature as a covariate for movements upstream and into
tributaries, but not downstream.
- Spill
- March days of spill
- One option is to use this as a covariate for overshoot fallback
only; might need to do some clever indexing with origins to make this
happen
- Flow - DROP. Flow is highly correlated with spill, and spill has
been found to be more informative than flow for downstream movements
(fallback) and not a good predictor of upstream movements
- Richins and Skalski (2018) investigated how fallback rates were
related to flow, and found no relationship between fallback rates to
natal tributaries and late-winter flow
- Siegel et al. (2021) looked at upstream migration of Steelhead
(between Bonneville and McNary) and found that flow was not a good
predictor for upstream migratory delays
Categorical
- Natal origin - all movements within ESU boundaries
- Hatchery vs. wild - all movements
- Barged vs. not barged - all movements
- Note that Richins and Skalski (2018) found that barging led to lower
migration success (but also lower overshooting), but only had enough
data for Tucannon River Hatchery Steelhead
- Acclimated vs. not acclimated - all movements
- Ocean age - all movements
Inevitably we are going to have fish where we are using covariates
where we know they didn’t experience those conditions. If its only off
by a few days it shouldn’t be a big deal because conditions are
generally very autocorrelated, especially in the case of things like
temperature.
Residence time by mainstem reach
Covariate data exploratory analysis
What is the relationship between water temperature at Ice Harbor and
the probability of overshooting (and other movements) for Walla Walla
River Steelhead?
From Richins and Skalski (2018): for natural-origin Walla Walla River
steelhead, the probability of moving directly to home decreased from
over 90% to less than 25% as water temperatures increased from 10°C to
20°C.

run the multinomial logit again with fish, but now only those going
upstream

run the multinomial logit again with fish, but now only those going
downstream
